IT Home: SuGuan 8000 Supports Reusable Rocket Research

Chinese technology outlet IT Home reports that SuGuan 8000, a domestically produced AI supercluster, was used by a research team from Xi'an Jiaotong University to complete ultra-large-scale, high-precision numerical simulations of rocket exhaust plumes. The project achieved a new international record for high-precision simulation of supersonic spatio-temporal coupled jets, filling a gap in domestic high-performance computing platforms for large-scale simulation of multi-nozzle supersonic jet flows.

The interference between exhaust plumes from multiple engines is a critical issue in the development of reusable rockets. When multiple engines operate in parallel, their high-temperature, high-velocity exhaust jets can interact, creating complex pressure fluctuations and heat distributions. Due to extreme environmental constraints, obtaining complete internal flow field information through physical experiments is difficult, making high-precision numerical simulation a vital research method.

The research team performed a 3D high-precision simulation of a 33-nozzle annular array on the SuGuan 8000. This recreated the complex processes of merging, interference, and mixing of multiple exhaust plumes, akin to an "ultra-high-definition CT scan" of the engine cluster's tail flame. The results are expected to aid in rocket body design, thermal protection optimization, and subsequent experimental planning, thereby advancing reusable rocket development.

The simulation pushed the SuGuan 8000's capabilities, involving 6.774 trillion grid cells with a weak scaling efficiency of 99.01% and strong scaling efficiency of 73%. The grid update throughput reached 34.029 trillion cells per second, validating the system's ability to support ultra-large-scale parallel computing. The system's high-speed interconnect and efficient storage ensured the stable and efficient execution of the complex task, paving the way for future larger and more precise studies.